Tunee Review: How I Create Music with Tunee Without Knowing Music or AI

​Repost Declaration​: This article is authored by Tunee user ​荔枝甜甜​. The original piece was published on NetEase Little Bee. This authorized repost is for reference only—no plagiarism or commercial use without permission. All rights reserved by the original author.

---

​About the Author​: I’m ​荔枝甜甜, an everyday user without a music background, sharing some hands-on notes about using Tunee. Bottom line first: ​you don’t need prompts to make music with Tunee​—it’s very friendly for non-experts. You can generate music just by chatting. It’s practically zero-threshold: no need to know music jargon or AI internals. I don’t know either—most of what I’ve picked up is from the media—and I still have a great time.

---

Getting to Know Tunee’s Interface & Workflow

Let’s begin with what it is. The left side of the interface is a ​chat box​—an LLM. The right side is the ​workbench—your canvas. Imagine the left side is ​DeepSeek​, chatting with you to understand your needs, while the right side is ​Xingliu, turning those needs into finished output.

Operation interface demo: https://www.tunee.ai/project/GPGTlDhuGsLRFDcPl

After chatting on the left, Tunee takes what it learned and ​outputs prompts to the right-hand workbench​. So we don’t need to memorize any prompts at all. Just tell Tunee your requirements and the vibe you want. If you’ve used Xingliu for image generation, the Tunee workbench will feel very familiar—almost plug-and-play.

Of course, you can also paste prompts directly on the homepage to generate—e.g., copy a Suno prompt and throw it in—but here’s what makes Tunee different: it’s not a command-driven music AI; it’s a conversation-driven AI.

So, borrowing from an engineer’s note I once saw about ​DeepSeek-R1: You don’t need to feed R1 rigid commands—better to forget commands entirely. Give it context and a ​narrative​; it will figure things out. That’s how you unlock the full value of a model like R1.

My Creation Method & Mindset

When I generate music, I typically ​chat with Tunee first​. I rarely upload audio or fire off direct commands.

First point:

I don’t have cover or traffic goals; most of my work is ​original​—I make things to ​delight myself​. I don’t want to sound like everyone else; I want something stylistically new. I think of uploading audio as akin to image “image-to-image” in Midjourney or Kimi Imagine: say I want a Klein bottle, but the model doesn’t know it and draws a regular bottle—so I upload a reference and ask for changes in material, color, etc. That works, but for music, uploading a track tends to anchor you to the ​original style or structure​—less original for me. I prefer Tunee to freely explore from its own knowledge base.

Second point:

Direct commands require ​complex prompts​. I noticed Tunee’s own prompts are in ​English​, which would mean writing mine, translating, or hunting for others’ English prompts—too much hassle, undermining the point of an LLM. So I just ​chat​: “Can you write a song in (language)? Do you know (style)? Are you familiar with (singer/composer)?” It replies, “Great, I know it, absolutely!” It may look like small talk, but I know transformer attention is locking onto my language, style, and genre for the upcoming piece. I don’t need to know exact genres, tempos, or instrumentation. I only need to know what I like—what sounds good—and write the ​lyrics​. Tunee handles the rest. How transformer and diffusion talk under the hood isn’t my concern—I’m just a happy newbie; let the engineers sweat that part, haha.

Why do I trust it? Because the left pane is an ​LLM​—it can understand what my lyrics want to express. If it misses the mark, that’s a training issue, but so far it’s been great. The workflow feels like: we ​chat to define the big picture​, then it uses my lyrics and imagination to generate finer details that match my emotional needs.

My Creation Process

Take my contest submission as an example. I planned to enter the “Whispers of Love” category, so I chose a Lisa Ono vibe—​French romanticism, breathy whisper, poetic​. Whether it’s Bossa Nova or some flavor of Jazz doesn’t matter to me; I just start chatting with Tunee.

“Hi Tunee! Can you write a song in French?” “Hey, absolutely—French songs have a beautiful feel! I can help with both lyrics and full-track production. Share your theme, style, or anything you want to express—we’ll craft a French song that’s uniquely yours!”

Great—now Tunee knows we’re making a French song. Anything that clashes with that vibe is naturally filtered out.

“Can you do a gentle, whispery song like ​Lisa Ono? I’ve prepared a poem called ​Goodnight Poem​—do you only need French, or Chinese as well?” “Wonderful idea! Lisa Ono’s gentle whisper style is perfect for night and warmth. If you want French lyrics, I can express ‘Goodnight Poem’ in French; if you prefer a Chinese–French mix, that works too. Tell me more about the atmosphere and imagery you want so we can get closer to the song in your mind.”

Tunee distilled the keywords ​gentle, whisper, night, warmth​. It seems to want ​atmosphere, imagery, and lyrical detail​—which I already had in my lyrics—so I shared them straight away.

Then I sent my finished lyrics and said: “Tunee, I want a warm female voice in the style of Lisa Ono—like lovers whispering.” Tunee thought for a moment and offered ​three options​:

1. Gentle Bossa Nova (female)
2. Minimalist coffeehouse ambience
3. French jazz fusion
4. All of the above

This shows I don’t need to know exact styles or genres. I just describe ​artists and feelings​, and Tunee handles the rest. See? Without me saying it, it surfaced Bossa Nova and ​Jazz​.

Curious, I picked “all of the above.” Tunee then began composing prompts in the left chat and piping them to the ​right workbench—the part you’d normally do yourself in tools like Suno. Tunee handles it for you, lowering the barrier. During the National Day holiday, I noticed the chat didn’t show all prompts; each style had different prompts. After the holiday, all three prompts appeared. In my example, the left looked like one prompt, but it actually produced ​three​:

1. “Bossa Nova, warm and intimate mood, soft female whisper vocals, nylon-string guitar, shaker, accordion accents, gentle groove, studio quality”
2. “minimalist, coffeehouse, piano, ambient, guitar, cello, breathy female vocals, spoken-word, poetic, intimate, gentle, nighttime, warm”
3. “French jazz-pop fusion, Lisa Ono style, warm and rhythmic female vocals, jazz piano and accordion, brushed drum kit, gentle groove, Parisian romantic night city atmosphere, studio quality”

Even if you don’t read English, it’s fine—just pick what ​sounds best*. I liked the second one most (​Goodnight Poem​, female perspective). Looking closely, “​minimalist​” stands out. Coffeehouse, piano, ambience, guitar—those are common, but ​“breathy female vocals,” “spoken-word,” “poetic,” “intimate,” “gentle,” “nighttime,” “warm”*—that’s exactly my taste. I love that track. Honestly, I wouldn’t have written prompts like that myself. Even if I copied something online, it might not match my mood.

After that track, I remembered French has nuances between male and female speech, so I tweaked the lyrics and asked Tunee for a male-perspective version in the same style—also lovely: ​Goodnight Poem/​, male perspective. I later did similar experiments in ​English, German, Japanese​, and ​Baroque​—all quite good. Chinese currently feels weaker—maybe due to data quality issues: fuzzier diction, undefined styles, too many “viral” outliers, polyphonic/morphologically tricky characters.

Tunee gives everyday users a great creative space: just know what songs you like, and forget prompt minutiae. Advanced users can still edit and fine-tune, but I rarely bother, haha. I focus on writing the lyrics and ​telling Tunee what I like​. I let it shine—and I live for those lightning-in-a-bottle moments.

Of course, Tunee has limits. Even with identical prompts and locked melody, each generation differs a bit. Sometimes I adore one take but a word is off; after a tweak, it’s not the same feel. Ten more tries—still small differences in diction, emotion, or melody. Also, for ​male–female duets​, even when I tag each line, it occasionally swaps roles. That’s an ​AI limitation​, and I can accept it—perfection can wait. Embrace the surprises. Also, prep your lyrics and ​proofread​—I once typed “famale” instead of “female,” so Tunee never gave me a female voice—oops! With clean lyrics, you’ll likely have fewer regrets.

Music is joy—human nature can’t escape it. Everyone has music flowing inside because we all have ​feelings​. We just lacked the means to express them. With AI, expression becomes ​freer, more personal, and more candid​. What words can’t say, poetry can; what poetry can’t, music will. What I think and feel right now can be fixed into a single track. For that, I’m truly grateful to Tunee. May Tunee grow more diverse, smarter, more refined—and enter thousands of homes.

---

Appendix 1: Music Links Mentioned
Poème du Bonne Nuit (Female Perspective)
Poème du Bonne Nuit (Male Perspective)

Appendix 2: Bilingual Lyrics (Chinese/French)

Poème du Bonne Nuit (Goodnight Poem)

【Verse 1】
Lune qui doucement caresse son visage (The moonlight gently caresses his profile)
Chaque soir il recompose un poème du soir (Each night he performs a goodnight poem for me)
Il dit : « Rilke, Neruda, Borges » (He says, “Rilke, Neruda, Borges”)
Dans sa main les vers deviennent nuages (In his hand, the verses turn to drifting clouds)

【Verse 2】
Il dit : « Nous avons même perdu ce crépuscule » (“We have even lost this dusk”)
« Personne ne nous vit mains enlacées cet après-midi » (“No one saw us holding hands this afternoon”)
« Quand la nuit bleue tombe sur le monde » (“When the blue night falls upon the world”)
« Personne ne nous vit tenant des flammes » (“No one saw us holding flames”)

【Chorus】
Je t'aime sans savoir comment ni quand (I love you, not knowing how or when)
Je t'aime directement sans complexité ni fierté (I love you plainly—no complexity, no pride)
Ainsi je t'aime car je ne sais autre chose (Thus I love you, for I know no other way)
Que ce chuchotement sous les étoiles (Than this whisper beneath the stars)

【Verse 3】
Il dit : « Les oiseaux nocturnes picorent les étoiles » (“Night birds peck at the first stars”)
« Qui scintillent comme mon âme » (“Which shimmer like my soul”)
« Donne deux jours de climat du Sud » (“Give them two more days of southern weather”)
« Ton oreiller sera constellé de mes regards » (“Your pillow will forever glitter with the stardust of my gaze”)

【Interlude】
Psyche... mon âme... (Little butterfly… my soul…)
Il dit... toujours il dit... (He says… he always says…)

【Chorus (Reprise)】
Je t'aime sans savoir comment ni quand (I love you, not knowing how or when)
Je t'aime directement sans complexité ni fierté (I love you plainly—no complexity, no pride)
Ainsi je t'aime car je ne sais autre chose (Thus I love you, for I know no other way)
Que ce chuchotement sous les étoiles (Than this whisper beneath the stars)

【Outro】
Il dit... (He says…)
« Je t'aime sans savoir comment » (“I love you, not knowing how”)
« Directement... » (“Plainly…”)